Real time monitoring of driver’s attention and fatigue typically is approached in controlled conditions, which cannot be valid for real-world applications. This work focuses on the use of deep learning to classify the inattentive states providing a classification model independent from a specific training for each single user. The database for the experiment is collected through a camera mounted on the car dash, recording live video with adverse light conditions. The proposed system proved to have high accuracy at a low computational cost with a fast processing time, viable for real time car applications, and tested in a ground-truth environment, far from the favourable controlled conditions of a simulator. Moreover, this approach offers an improvement in the classification stage, being error-free when pre-training the system in the detection of fatigue and inattention of a specific user and, with minimal error on a random user dataset.
Detection of driver’s inattention: a real-time deep learning approach
2019-10-01
449594 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
METHOD AND APPARATUS FOR JUDGING DRIVER'S INATTENTION
Europäisches Patentamt | 2015
|Springer Verlag | 2017
|Real-Time Driver's Stress Event Detection
Online Contents | 2012
|Real-time driver's eye state detection
IEEE | 2005
|Real-Time Driver's Stress Event Detection
IEEE | 2012
|